Medium

ONNX: Out-of-Bounds Read in RNN/GRU/LSTM Shape Inference (Old Opsets, Missing Rank Check)

  • Disclosed: 28.10.2026
  • Updated: 28.10.2026 23:17

Summary

ONNX shape inference performed an out-of-bounds read when processing a malformed RNN, GRU, or LSTM node from opsets 1–6 whose first input had a rank other than three.

RNNShapeInference_opset1_to_6() checked whether the input shape was present but accessed dim(0) and dim(1) without first checking its rank. A crafted model could therefore terminate a process that called onnx.shape_inference.infer_shapes(). The model passed onnx.checker.check_model() before reaching the vulnerable path.

Technical details

The vulnerable code read dimensions unconditionally:

cpp
if (hasInputShape(ctx, 0)) {
  const auto& first_input_shape = getInputShape(ctx, 0);
  seq_length = first_input_shape.dim(0);
  batch_size = first_input_shape.dim(1);
}

For an undersized shape, protobuf's repeated-field access read outside the intended array in release builds. The observed result was SIGSEGV; sanitizer and debug builds terminated on a protobuf bounds check.

The newer opset 7–13 and opset 14+ inference paths already required the first input to have rank three and were not affected.

Impact

An application that performs shape inference on an untrusted model can be terminated, causing denial of service. Exploitation does not require privileges or human interaction once the model is processed. No confidentiality or integrity impact was demonstrated.

Affected versions

  • Affected: ONNX through 1.21.0
  • Patched: ONNX 1.22.0

The vulnerable schemas are limited to RNN and LSTM opsets 1–6 and GRU opsets 3–6.

Remediation

Upgrade to ONNX 1.22.0 or later. The fix validates that the first input has rank three before accessing its dimensions.

The original malformed model was also re-tested against the published ONNX 1.23.2 wheel and correctly raises InferenceError.

Severity

The vulnerable ONNX component processes the malicious model through a local library API or file boundary and is not itself bound to a network stack. Applications that expose ONNX model processing over a network should calculate an Environmental score with Modified Attack Vector set to Network.

Affected software

Click a technology to see all vulnerabilities affecting it.

PlatformPackageVulnerable versionsFixed in
Python (PyPI) onnx < 1.22.0 1.22.0

References

CVSS score 6.2/10
CVSS vector
CVSS:3.1/AV:L/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H
Weakness (CWE)
CWE-125
Platform
Python (PyPI)
Package
onnx
Disclosed
28.10.2026
Updated
28.10.2026 23:17

Sources and licenses

OSV / GitHub Advisory Database - GHSA-48m4-24cq-37qf - CC BY 4.0

Vulnerability history: onnx

All published vulnerabilities of this package (Python (PyPI)), newest first. Shows whether this is a one-off slip or a package that needs watching.

All (17) in the archive
Total entries
17
First → latest
01.2023 10.2026
Without a fix
0
Exploited in attacks
0

Severity breakdown

  • Critical 1
  • High 7
  • Medium 8
  • Low 1

Disclosures per year

  1. 2023: 1
  2. 2024: 3
  3. 2025: 1
  4. 2026: 12
Severity:
  1. ONNX: Out-of-Bounds Read in RNN/GRU/LSTM Shape Inference (Old Opsets, Missing Rank Check) this entry
    Medium · 6.2 CVE-2026-115919 < 1.22.0 1.22.0
  2. High · 7.1 CVE-2026-49114 < 1.21.0 1.21.0
  3. High CVE-2026-27489 < 1.21.0 1.21.0
  4. High · 7.5 CVE-2024-27318 < 1.16.0 1.16.0

Know which vulnerabilities affect your application

With the DockRay integration and software inventory enabled, DockRay checks reported component versions against known vulnerabilities and notifies you about confirmed matches. Coverage depends on the supported technology and available advisory data.

Connect your application
DockRay newsletter

Application security in your inbox

Important vulnerabilities, actionable guidance and practical advice for your applications. Choose your technologies and follow what matters to your project.

  • A daily digest of new vulnerabilities, critical ones included.
  • Notifications about important updates.
  • Practical advice on application security and monitoring.

Leave empty to receive updates for every technology.

Free to subscribe. Unsubscribe with one click in any email.